Latest / StartUp Health NOW Podcast / Nigam Shah MBBS, PhD, Stanford University

Transcript

Nigam Shah MBBS, PhD, Stanford University · StartUp Health NOW Podcast

• 0:00
  [on-hold music] We are with Dr. Shah, the founder of Cardinal Analytics. How are you, sir? Good. So first, um, can you give us, uh, paint the vision for Cardinal Analytics?
• 0:14
  Tell us, tell us your plan and what you're doing.
• 0:16
  Actually, the story began a couple of years ago, uh, when, uh, my co-founder, Arnie Milstein, and I, uh, met with Tom McKinley, who's one of the investors now, uh, on what could we do to change the value of care by using machine learning.
• 0:30
  And we looked around, and we saw a lot of things where people were making predictions, and then you kind of email the list of predictions to the CEO, and you raise their blood pressure, but very little focus on what you could do about the prediction that you just made.
• 0:44
  So we focused on both arms, so that you make a prediction, but along with that, you also recommend an action that somebody could take in order to prevent the bad outcome that you might have predicted.
• 0:55
  So that's sort of the founding DNA, uh, and idea behind the company that you have to make a prediction, but along with the prediction, you also recommend some concrete action- Action...
• 1:05
  that someone can take in order to mitigate the, a risk that you just predicted. The first challenge is there's all this extraordinary new data, right? The second challenge is what do we do with that data- Mm-hmm...
• 1:17
  to actually have predictions that, that are accurate and precise. And what you're saying is, okay, with the prediction now, what are we gonna do with that? Exactly. Turn it into something actionable.
• 1:29
  Um, very interesting. So what are your thoughts on, um, all the new data that's coming in, um, to the market really? Mm-hmm. Um, there's trillions of sensors now.
• 1:43
  There's data around us, in us, you know, in all these great devices. Um, what does that mean for, for what you're doing at Cardinal, and how do you think about sort of this wave and influx of data that's coming in?
• 1:58
  So new data is always an opportunity, but you have to ask yourself the question that does these data or these new data that are coming allow me to predict that I couldn't have otherwise predicted?
• 2:11
  And most important, do they give me any way of taking an alternative behavior that would mitigate the prediction? 'Cause predictions are dime a dozen. Action is priceless. Hmm.
• 2:23
  And to the extent that your prediction does not change action, it's useless, and to the extent it does, that's how val- as valuable as it gets. So would you rather have more data or, or less?
• 2:34
  I would rather have less data that gives me actionable insight, as opposed to more data that allows me to make obvious predictions. Right. So I can make an obvious prediction. You're gonna die in the next hundred years.
• 2:42
  What are you gonna do? [laughs] Right. It's useless. Yeah. And there's so much such, that kind of predictions going around. I mean, risk stratification, scoring, where the recipient of the prediction already knows,
• 2:56
  and by giving them a number, yes, you quantify the risk, but you don't help them in mitigating the risk. So, so with your solution, what are some tangible examples of what people may be able to, to actionably do?
• 3:10
  So one of the things that we focus on is that of the people that we predict are at risk, for the ones that are at the most risk, we actually have a team of doctors that review their prior medical record, and based on their experience, take an educated guess as to what intervention that if we did for them, would help in preventing the risk that we're predicting.
• 3:34
  Hmm. And that's the key difference, that most of the times an analytics company gives you software that you use and create a dashboard,
• 3:43
  and then it gives you a list of people who are gonna get into trouble in red, yellow, green or whatever. Uh, we're not selling software. Right. We're selling care plans.
• 3:53
  As in the goal is to take, lead you to the person on whom if you intervene, and in a specific way, you would actually improve the quality of care for that person. So who's your customer today?
• 4:07
  Predominantly people or entities that are at com- full risk, so it'll be health plans to start with, uh, fully integrated payer provider systems are obviously a place, uh, where this has a role.
• 4:18
  Um, anyone who's faced with the issue of, I have a thousand people or ten thousand people. I have the capacity to take action on ten of them or a hundred of them. Which are the hundred that I call today?
• 4:31
  We answer that question, but we also tell you what do you say when you call them. Right. So there's been a, an influx of capital coming into analytics in around health and healthcare over the past few years.
• 4:43
  Um, I think ultimately that's a good thing because as we talked about, there's all this data. Um, there needs to be proper analysis of it- Mm-hmm... so that it can then be actionable. Um,
• 4:54
  how is what you're doing different than, than what you're, you're seeing with a wave of other analytics platforms that are in the market today? So I mean, I'm not an expert on this sector, so, uh- Yeah....
• 5:07
  don't quote me on this on Wall Street. Uh, but the typical analytics company is about taking data, making predictions, and creating a beautiful report about whatever it is that they're trying to predict.
• 5:21
  That is not the goal for us. The goal is once you have a prediction, let's look at the individual, the person who's at risk, and devise a customized solution that works for them. Hmm.
• 5:32
  So it's the action part that takes priority, and you see that even in the two founders.
• 5:37
  I'm the analytics co-founder, and Arnie Milstein, who's a nationally known person in, uh, uh, providing good value care, he's the, uh, other co-founder.
• 5:45
  So does the provider then get access to that or the patient or both? Uh, right now it's through the health insurance plan and their medical management teams. Okay. Uh, in the future, it could be providers. Um,
• 5:59
  not patients in the immediate short term because the kinds of interventions that are needed is, for example, to get somebody a health coach or to find them an expert in GI care.
• 6:09
  I mean, that's not something a patient can do for themselves. The health system has to take that action on their behalf.So one of the challenges we've come across is with a lot of the analy- analytics platforms is
• 6:21
  the organizations using them need to then hire a, a wave of data scientists to actually understand- Exactly... what's going on. How... Tell me about that and how that works with, with what you're doing at Cardinal.
• 6:34
  So we're not giving them a data science platform, so they don't have to hire a whole bunch of data scientists.
• 6:39
  The interface is with the people at health plans who are already s- tasked with managing the care of individuals. They're typically called care management teams or disease management teams. Mm-hmm.
• 6:51
  They're the recipients, and what they- So they can log in and start using it and- In fact, the care plan right now goes as a, as a, as a, a portal to them, but in the future can integrate with whatever case management system they're already using.
• 7:04
  Mm-hmm.
• 7:05
  And so what they get is this nugget of recommendations saying, "For John Smith, uh, we believe you should switch this drug to this drug, and we believe that they should get a, a visit with this kind of a GI expert."
• 7:16
  Mm-hmm. And those are the actionable items that go to them, and they come from a board-certified MD. So what are you most excited about
• 7:25
  the next few years in terms of what we're gonna be able to do with platforms like yours? So imagine collecting this data for
• 7:34
  four years, that you've made a prediction, a machine, a black box is saying, "This person will get into a tr- in- into a problem." And then a human looking at that person's record and recommending an intervention.
• 7:46
  You do that for four years, you create a unique data set from which you can actually learn what actions are useful and effective, and right now we don't have that, and collecting that data in a prospective setting and train the next generation of learning algorithms that can even recommend a use- useful intervention is what I'm most excited about.
• 8:07
  So where can people go to learn more about Cardinal Analytics? So there's our website, cardinalanalytics.com.
• 8:12
  That's the place to start, and, uh, our other members are here, uh, our CEO, uh, Tom McKinley, and the, uh, VP of Sales Afsan Akthar, and, uh, could meet with them. Excellent.
• 8:23
  So we'll, uh, we'll put those links in the show notes, and, uh, thank you for everything you're doing. All right. Thank you.